search

neonbjb / tortoise-tts

A multi-voice TTS system trained with an emphasis on quality
Apache License 2.0
12.33k stars 1.72k forks source link

Is additional training possible with tortoise-tts? #738

Open kbuwel opened 4 months ago

kbuwel commented 4 months ago

I am very interested in this project. I think it's an interesting project that can create tts with a 10-second voice sample. I also think it's good to support multiple languages. However, there is a pronunciation problem. I've thought for a long time that if only the pronunciation is corrected, it could be a very good tool. Tortoise-tts read me a document written in Korean, but it's not very good to listen to in English pronunciation. I had the thought that I wanted to add Korean pronunciation to tortoise-tts. I want to know if additional training is possible and if there is a license cost, I am willing to pay if it is a reasonable cost. I would appreciate it if you could tell me how to add training to tortoise-tts.

manmay-nakhashi commented 4 months ago

@kbuwel it's an open source project , no licensing required, you need to train a model from scratch for a new language with lots of data > 10k hours more is better, train a bpe tokenizer for Korean language.

kbuwel commented 4 months ago

Is it possible to create tts with Korean pronunciation using just a tokenizer? Don't need audio data or large-text transcripts?

manmay-nakhashi commented 4 months ago

@kbuwel no

dhamaraiselvi commented 1 month ago

I need documentation to train the model for Tamil and Indian English with tortoise-tts. No proper information to build from scratch